Problem Statement

The least common multiple (denoted "lcm") of a non-empty sequence of positive integers is the smallest positive integer that is divisible by each of them. For example, lcm(2)=2, lcm(4,6)=12, and lcm(1,2,3,4,5)=60.

Alice had a positive integer N. Then she chose some positive integer M that was strictly greater than N. Afterwards, she computed two values:

the value A = lcm(N+1, N+2,
..., M) and the value B = lcm(1, 2, ..., M). She was surprised when she saw that A = B.the value A = lcm(N+1, N+2,
..., M) and the value B = lcm(1, 2, ..., M). She was surprised when she saw that A = B.the value A = lcm(N+1, N+2,
..., M) and the value B = lcm(1, 2, ..., M). She was surprised when she saw that A = B.the value A = lcm(N+1, N+2,
..., M) and the value B = lcm(1, 2, ..., M). She was surprised when she saw that A = B.the value A = lcm(N+1, N+2,
..., M) and the value B = lcm(1, 2, ..., M).

You are given the int N. Find and return the smallest M Alice could have chosen. (Such an M will always exist.)

Definition

  • ClassMissingLCM
  • MethodgetMin
  • Parametersint
  • Returnsint
  • Method signatureint getMin(int N)
(be sure your method is public)

Limits

  • Time limit (s)2.000
  • Memory limit (MB)256

Constraints

  • N will be between 1 and 1,000,000, inclusive.

Test cases

    • N1

    Returns2

    Alice needs to choose an M > 1 such that lcm(2,...,M) = lcm(1,...,M). We can see M=2 is the minimum value that works, since lcm(1,2) = lcm(2) = 2.
    • N2

    Returns4

    • N3

    Returns6

    We have lcm(4,5,6) = lcm(1,2,3,4,5,6) = 60.
    • N4

    Returns8

    • N5

    Returns10

    • N42

    Returns82

    Oh... that doesn't fit the pattern.
      • N999999

      Returns1999966

    +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

  1. In this problem the actual LCM values used can be notably large. It is best to avoid approaches that calculate them directly. Then how about we think of this problem in terms of the prime factorization of the numbers. For example, consider two numbers: 12 and
    135. Their prime factorizations are: 22⋅3 and 33⋅5.
    The prime factorization of the LCM will be: 22⋅33⋅5.
    In other words, for each prime number , we take the maximum exponent of the prime number among the two numbers we are calculating the LCM for. How does this translate to the two LCMs used?

    A=LCM(N+1,N+2,...,M) 

    B=LCM(1,2,...,M) 

    A=B

    We need to translate this into distinct conditions, one for each relevant prime number:

    a=max(ep(N+1),ep(N+2),...,ep(M)) 

    b=max(ep(1),ep(2),...,ep(M)) 

    a=b

    Where ep(x) is
    the exponent of prime number p in
    the prime factorization of x,
    the maximum number of times you can repeatedly divide x by p.
    In words we just want the maximum exponent of prime p among
    all the factorizations between N+1 and M and
    between 1 and M to
    be equal. Try this:

    b=max(ep(1),ep(2),...,ep(N),ep(N+1),...,ep(M)) 

    b=max(ep(1),ep(2),...,ep(N),max(ep(N+1),...,ep(M))) 

    b=max(ep(1),ep(2),...,ep(N),a)

    What happens here is we take advantage that the maximum operation is associative. This means max(a,b,c)=max(a,max(b,c))=max(max(a,b),c).
    The useful thing to conclude about this: b≥a.
    There's more:

    b=max(max(ep(1),ep(2),...,ep(N)),a) 

    c=max(ep(1),ep(2),...,ep(N)) 

    b=max(c,a)

    We want a=b=max(c,a):
    This means we want c≤a.
    Note that c is
    constant, as it's determined by the numbers between 1 and N.
    So we just need to look for a value of M such
    that the maximum exponent of p among N+1,N+2,...M is
    greater than or equal to c.

    There must be a number x>N for
    which ep(x)≥c,
    this means that the exponent of p in
    the prime factorization of x.
    If we take the maximumep(i) for
    all i between N+1,...,M,
    the result would be at least c,
    meaning that using M=x would
    be correct. M=x+1 and
    any Mgreater
    than x would
    also be correct. If we find the minimum M that
    is valid for p,
    then we can assume that all greater numbers will also be valid for p.
    This minimum M will
    be the smallest x>N for
    which ep(x)≥c.

    Did you notice that in the examples it initially appears that the result is always 2N and
    then the result seems to be smaller than 2N but
    not far apart?

    There is a good explanation for this. c is
    the maximum exponent of p for
    some number less than or equal to N,
    let's call that number m. 2m will
    also have that exponent for p (unless p=2,
    in which it will have an even larger exponent). Making 2m a
    valid value for M.
    If we pick M=2N,
    we will guarantee that this happens for all relevant prime numbers. This is useful because it means we only need to search for xamong
    numbers less than or equal to 2N.

    For each prime p,
    there will be a distinct minimum valid M ,
    we should pick the maximum out of all of them. This number will be valid for all primes we try. Note that when p>N,
    any M>N will
    be correct. Because when p>N , c is
    zero, none of the numbers smaller than or equal to N will
    be multiples of p,
    so the maximum exponent will be 0. This means we only need to repeat this for all the prime numbers that are less than or equal to N.

    For each prime number we need to find c and
    also find the minimum x such
    that: ep(x)>c and N<x≤2N.
    The final improvement we need is to notice that given p,
    we only need to think in terms of numbers that are multiples of p.
    For i≤N,
    only values of i that
    are multiples of pwill
    have an exponent of p in
    their prime factor, so only they are relevant for the calculation of c.
    For i>N,
    only multiples of p may
    have an exponent larger than or equal to c.
    In total we will try all the multiples of p less
    than or equal to 2N.
    This is repeated for each p<N.
    The final number of steps needed is: 2N2+2N3+2N5+...+2NP where P is
    the maximum prime that doesn't exceed N.
    This number of steps is very good and the complexity would be similar to the Sieve of Eratosthenes'. A simple way to tell that complexity is O(NN−−√) :
    For all P>2N−−−√, 2NP=1.
    There are 2N−2N−−−√ such
    values, this is O(N).
    For the other O(N−−√) values
    of p,
    even if we assumed O(N) steps,
    the total complexity would still be: O(N+NN−−√).
    We are ready to implement this solution:

    vector<int> get_primes(int N)
    {
    // Sieve of Erathostenes to find all the necessary prime numbers:
    vector<int> res;
    vector<bool> composite(N + 1, false); for (int p = 2; p <= N; p++) {
    if (! composite[p]) {
    for (int i = p+p; i <= N; i += p) {
    composite[i] = true;
    }
    res.push_back(p);
    }
    }
    return res; } int get_exponent(int x, int p)
    {
    int r = 0;
    while (x % p == 0) {
    r++;
    x /= p;
    }
    return r;
    } int getMin(int N)
    {
    int res = 2;
    // For each prime number <= N:
    for (int p: get_primes(N) ) {
    // first find the maximum exponent of p among numbers <= N
    // (in the explanation , this max_exp is c)
    int max_exp = 0;
    int i = p;
    while (i <= N) {
    max_exp = std::max(max_exp, get_exponent(i,p) );
    i += p;
    }
    // seek the minimum i such that get_exponent(i,p) >= max_exp:
    while (get_exponent(i,p) < max_exp) {
    i += p;
    }
    // the maximum for all ps is the result:
    res = std::max(res, i);
    }
    return res;
    }

最小公倍数 SRM 661 Div1 250: MissingLCM的更多相关文章

  1. Topcoder SRM 643 Div1 250<peter_pan>

    Topcoder SRM 643 Div1 250 Problem 给一个整数N,再给一个vector<long long>v; N可以表示成若干个素数的乘积,N=p0*p1*p2*... ...

  2. topcoder srm 661 div1

    problem1 link $N+1$到$M$ 之间的数字要包含所有1到$N$之间出现的质因子的最高幂即可. problem2 link 从第一个节点到第$N$个节点依次考虑.对于第$i$个节点来说, ...

  3. SRM 595 DIV1 250

    挺简单的组合把. #include <cstdio> #include <cstring> #include <iostream> #include <vec ...

  4. SRM 594 DIV1 250

    可能开始宿舍比较乱,思绪静不下来...想了大半个小时,终于确定了应该暴力+DP,然后写了枚举除数,和被除的版本..这样,还敲错了个字母,第一次提交还100多,修改提交还有75.多,最后想到,貌似不打对 ...

  5. TC SRM 593 DIV1 250

    我只能说的亏没做,要不就挂0了.. 本来想四色定理,肯定4就可以的...然后准备爆,发现3的时候不好爆,又想了老一会,嗯,数据范围不小,应该不是暴力,直接找规律,貌似最大就是3,有一个3连块,输出3, ...

  6. TC SRM 593 DIV1 250(dfs)

    这图最多3色就可以 搜2就行了 #include <iostream> #include<cstdio> #include<cstring> #include< ...

  7. Topcoder SRM 698 Div1 250 RepeatString(dp)

    题意 [题目链接]这怎么发链接啊..... Sol 枚举一个断点,然后类似于LIS一样dp一波 这个边界条件有点迷啊..fst了两遍... #include<bits/stdc++.h> ...

  8. TopCoder SRM500 Div1 250 其他

    原文链接https://www.cnblogs.com/zhouzhendong/p/SRM500-250.html SRM500 Div1 250 题意 (看题用了半个小时--) 有 n 个人(编号 ...

  9. Topcoder Srm 726 Div1 Hard

    Topcoder Srm 726 Div1 Hard 解题思路: 问题可以看做一个二分图,左边一个点向右边一段区间连边,匹配了左边一个点就能获得对应的权值,最大化所得到的权值的和. 然后可以证明一个结 ...

随机推荐

  1. RPM软件包管理的查询功能 转

    RPM软件包管理的查询功能: 命令格式 rpm {-q|--query} [select-options] [query-options] RPM的查询功能是极为强大,是极为重要的功能之一:举几个常用 ...

  2. StatCounter

    StatCounter provides free customisable hit counters, visitor tracking, web analytics and website sta ...

  3. VS2008 LINK : fatal error LNK1104: cannot open file 'atls.lib'错误解决方案

    用VS 2008编写ATL的64位应用程序时,提示链接错误:VS2008 LINK : fatal error LNK1104: cannot open file 'atls.lib' 问题原因 VS ...

  4. AngularJS一个由于未声明对象而报的错

    实现这样的一个需求:点击某个按钮,然后显示或隐藏某块区域. 先注册一个AngularJS的一个module: var myApp = angular.module("myApp", ...

  5. JavaScript进阶系列07,鼠标事件

    鼠标事件有Keydown, Keyup, Keypress,但Keypress与Keydown和Keyup不同,如果按ctrl, shift, caps lock......等修饰键,不会触发Keyp ...

  6. JAVA BigDecimal的相加

    之前很少使用这样的一个对象BigDecimal,今天在改需求的时候遇到了,结果坑爹的怎么相加最后都为零. 代码如下: BigDecimal totalAmount = new BigDecimal(0 ...

  7. 在arcgis使用python脚本进行字段计算时是如何解决中文问题的

    来自:https://www.jb51.net/article/73561.htm 一.引言 在arcgis打开一个图层的属性表,可以对属性表的某个字段进行计算,但是在平常一般都是使用arcgis提供 ...

  8. poj 3041(最大匹配问题)

    http://poj.org/problem?id=3041 Asteroids Time Limit: 1000MS   Memory Limit: 65536K Total Submissions ...

  9. 找了一个api管理工具

    找了一个工具,https://github.com/nutsteam/apiManager选择了如下方式,进行了安装. ● 下载https://git.oschina.net/zhoujingjie/ ...

  10. 福尔摩斯基本演绎法第一季/全集Elementary迅雷下载

    本季Elementary Season 1 第一季(2012)看点:<福尔摩斯:基本演绎法>由CBS出品,根据<福尔摩斯>系列改编,地点从19世纪大英帝国国势鼎盛的时期搬到了2 ...